The main purpose of forensic DNA analysis is to perform human identification. However, in cases where no match is obtained, forensic intelligence leads may be generated from biological crime scene traces. Methods for prediction of externally visible characteristics such as hair-, skin- and eye-color or biogeographical ancestry are currently available and have successfully been implemented in our laboratory. Another characteristic of interest is the predicted age of a person that deposited a DNA trace.

Previous studies have established that DNA methylation levels in certain genes correlate with age, and several promising methods and models have been developed. 
In the last years, the focus has been on improving age estimates through finding alternative markers and adopting more sophisticated prediction models. However, age prediction methods have so far not been broadly implemented in casework laboratories, in part due to the lack of commercial kits, bioinformatic tools and open-source age prediction models.

Our aim is to adapt and implement methods for age estimation based on the analysis of crime scene samples. To this end, DNA methylation levels of age-correlated markers were measured by targeted sequencing of bisulphite-converted DNA using a MiSeq FGx instrument. In total, we validated two panels for blood and one panel for semen. These panels were previously developed by different research groups and consortia. The methods were modified to some extent from the original publications. The primer concentrations were adjusted for better marker balance, since drop-out occurred for samples of lower quality with the original protocols. Furthermore, a streamlined laboratory process was established by using a VeritiPro PCR instrument enabling different annealing temperatures within one run and applying the KAPA EvoPrep kit for effective library preparation for all panels.

We analyzed about 30 samples originating from blood from individuals in the range of 19 to 85 years, and 20 samples from semen from individuals 22 to 59 years. We investigated the limit of detection, matrix effects, repeatability and other relevant validation parameters. Additionally, the model performance and adjustments to improve the accuracy of the age prediction models were investigated. We will present the outcome of the validation study, including a comparison of the performance of the different panels and models. Lastly, we will discuss important lessons learned that can provide guidance for other laboratories aiming for implementation of age prediction for use in casework.

Authors

• Maja Sidstedt (National Forensic Centre, Swedish Police Authority, Sweden)
• David Juvén (National Forensic Centre, Swedish Police Authority, Sweden)
• Arvid Hedén Gynnå (National Forensic Centre, Swedish Police Authority, Sweden)
• Siri Aili Fagerholm (National Forensic Centre, Swedish Police Authority, Sweden)
• Johannes Hedman (National Forensic Centre, Swedish Police Authority and Division of Biotechnology and Applied Microbiology, Department of Process and Life Science Engineering, Lund University, Sweden)
• Ronny Hedell (National Forensic Centre, Swedish Police Authority, Sweden)